Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <2b935c82-ea5c-af47-4455-edf3a629e941@gmail.com>
Date:   Wed, 13 Apr 2022 15:03:13 -0700
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.7.0
Subject: Re: [PATCH net-next v3] net/ipv6: Introduce accept_unsolicited_na
 knob to implement router-side changes for RFC9131
Content-Language: en-US
To:     David Ahern <dsahern@kernel.org>, Arun Ajith S <aajith@arista.com>,
        netdev@vger.kernel.org
Cc:     davem@davemloft.net, linux-kernel@vger.kernel.org,
        linux-doc@vger.kernel.org, yoshfuji@linux-ipv6.org,
        kuba@kernel.org, pabeni@redhat.com, corbet@lwn.net,
        prestwoj@gmail.com, gilligan@arista.com, noureddine@arista.com,
        gk@arista.com
References: <20220413143434.527-1-aajith@arista.com>
 <5a92f5cd-9af4-4228-dc44-b0c363f30e18@gmail.com>
 <17769a5b-9569-18ee-d1c0-c8971a42c709@kernel.org>
From:   Eric Dumazet <eric.dumazet@gmail.com>
In-Reply-To: <17769a5b-9569-18ee-d1c0-c8971a42c709@kernel.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 4/13/22 15:00, David Ahern wrote:
> On 4/13/22 3:22 PM, Eric Dumazet wrote:
>> On 4/13/22 07:34, Arun Ajith S wrote:
>>> Add a new neighbour cache entry in STALE state for routers on receiving
>>> an unsolicited (gratuitous) neighbour advertisement with
>>> target link-layer-address option specified.
>>> This is similar to the arp_accept configuration for IPv4.
>>> A new sysctl endpoint is created to turn on this behaviour:
>>> /proc/sys/net/ipv6/conf/interface/accept_unsolicited_na.
>>
>> Do we really need to expose this to /proc/sys, for every interface added
>> in the host ?
>>
>> /proc files creations/deletion cost a lot in environments
>> adding/deleting netns very often.
> agree with the general intent (along with the increasing memory costs).
> I do think this case should be done as a /proc/sys entry for consistency
> with both ARP and existing related NA settings.
>
>> I would prefer using NETLINK attributes, a single recvmsg() syscall can
>> fetch/set hundreds of them.
> What do you have in mind here? A link attribute managed through `ip link
> set`?


Yes, something like that, if this is a netdevice/link attribute.